Azido inositol probes enable metabolic labeling of inositol-containing glycans and reveal an inositol importer in mycobacteria

Abstract

Bacteria from the genus Mycobacterium include pathogens that cause serious diseases in humans and remain difficult infectious agents to treat. Central to these challenges is the composition and organization of the mycobacterial cell envelope, which includes unique and complex glycans. Inositol is an essential metabolite for mycobacteria due to its presence in the structural core of the immunomodulatory cell envelope glycolipids phosphatidylinositol mannoside (PIM) and PIM-anchored lipomannan (LM) and lipoarabinomannan (LAM). Despite their importance to mycobacterial physiology and pathogenesis, many aspects of PIM, LM, and LAM construction and dynamics remain poorly understood. Recently, probes that allow metabolic labeling and detection of specific mycobacterial glycans have been developed to investigate cell envelope assembly and dynamics. However, these tools have been limited to structures in the mycobacterial cell envelope, including peptidoglycan, arabinogalactan, and mycolic acid-containing glycolipids. Herein, we report the development of synthetic azido inositol (InoAz) analogues as probes that can metabolically label PIMs, LM, and LAM in intact mycobacteria. Additionally, we leverage an InoAz probe to discover an inositol importer and catabolic pathway in Mycobacterium smegmatis. We anticipate that in the future, InoAz probes, in combination with bioorthogonal chemistry, will provide a valuable tool for investigating PIM, LM, and LAM biosynthesis, transport, and dynamics in diverse mycobacterial organisms.

Graphical Abstract

Introduction

Mycobacterial species remain a threat to human health and represent some of the most notorious human pathogens, including Mycobacterium tuberculosis, Mycobacterium leprae, Mycobacterium ulcerans, and a number of fast- and slow-growing non-tuberculous mycobacteria. Mycobacteria are collectively characterized by a glycan- and lipid-rich cell envelope that provides antibiotic tolerance and facilitates modulation of the host immune response.^{1, 2} Concomitantly, the unique composition and structure of the cell envelope of mycobacterial species is also a diagnostic marker and linchpin therapeutic target.^{3, 4} The distinctiveness of these biomolecules among bacteria and their distribution within the mycobacterial cell envelope are suggestive of highly specialized biosynthetic and transport machinery. Interrogation of cell envelope construction and dynamics and, specifically, elucidation of glycan biosynthetic and transport pathways are broadly important to understand the pathology of these microbes and to design therapeutic agents against them.

An important class of glycolipids in the mycobacterial cell envelope are phosphatidylinositol mannosides (PIMs) and their downstream products, lipomannan (LM) and lipoarabinomannan (LAM). PIMs, LM, and LAM are the subject of frequent studies due to their complex structures and critical roles in cellular integrity, permeability, and immunomodulation.^{1, 5–7} However, many proteins responsible for PIM/LM/LAM biosynthesis and transport are not known, nor is there knowledge of their dynamics in the cell envelope. The current summary of initial PIM biosynthesis involves the conjugation of one to six mannoses to a phosphatidylinositol (PI) core (Figure 1).^{1, 8} First, in the cytoplasm, PI is generated by the conjugation of D-myo-inositol-3-phosphate to cytidine diphosphate diacylglycerol (CDP-DAG) by the PI synthase PgsA1, followed by a dephosphorylation step catalyzed by an as yet unidentified phosphatase. PI is then mannosylated by GDP-mannose-dependent mannosyltransferases PimA and PimB’ at the C2 and C6 positions of inositol, respectively, to generate PIM₁ and PIM₂. Acylation of the C6 hydroxyl of the α−1,2-linked mannose of PIM₁ or PIM₂ is catalyzed by the cytoplasmic acyltransferase, PatA, to produce Ac₁PIM₁ and Ac₁PIM₂. From Ac₁PIM₂ onwards, the mannosyltransferase steps involved in the biosynthesis of Ac₁PIM₆ and LM/LAM are poorly defined. Although PimE has been demonstrated to transform Ac₁PIM₄ to Ac₁PIM₅ on the periplasmic face of the plasma membrane (PM), the protein(s) responsible for mannosylation of PIM₂ to PIM₄ and PIM₅ to PIM₆ remain to be identified. In addition, which form of PIM is translocated across the PM, and the protein responsible for this key transport step, are still unknown. Elongation of intermediate PIMs to higher PIMs, LM, and LAM on the periplasmic side of the PM involves glycosyltransferases that use the lipid-linked sugar donors decaprenylphospho-mannose (DPM) and decaprenylphospho-arabinose (DPA). In addition to being in the PM, PIM/LM/LAM are also present in the outer mycomembrane (MM), and the LAM-derived polysaccharides D-mannan and arabino-D-mannan have been identified in the capsule of M. tuberculosis.⁹ However, how these species are transported through the cell envelope to the outer layers is unknown. Additionally, the transporters responsible for the import of PIM/LM/LAM monosaccharide building blocks, including mannose and inositol, have not been previously identified. Finally, there is a lack of information regarding the spatiotemporal dynamics of PIM/LM/LAM construction and turnover during cell growth and division, which significantly lags behind knowledge of dynamics of other cell envelope components, such as peptidoglycan, arabinogalactan, and MM trehalose glycolipids.

Figure 1. Current knowledge regarding the biosynthesis and transport of PIMs, LM, and LAM.

PIM/LM/LAM synthesis is initiated on the cytoplasmic face of the plasma membrane by the addition of two Manp residues to phosphatidylinositol (PI) by dedicated, essential, GDP-mannose-dependent mannosyltransferases (PimA and PimB’). Di- or tri-acylated forms of phosphatidylinositol di- or tri-mannosides (PIM₂/Ac₁PIM₂ or PIM₃/Ac₁PIM₃) are then translocated to the periplasmic face of the plasma membrane by (an) as yet unknown inner membrane transporter(s). Elongation of these intermediate PIMs to higher PIMs, LM, and LAM on the periplasmic side of the plasma membrane involves glycosyltransferases (PimE, MptA, MptB, MptC, AftB, AftC, AftD, EmbC) that use lipid-linked sugar donors (decaprenylphospho-mannose and decaprenylphospho-arabinose). The transporters required to translocate the final PIM, LM and LAM products to the outer leaflet of the outer membrane are not known. See text for further details. AG, arabinogalactan; MM, mycomembrane; PG, peptidoglycan; PM, plasma membrane. Chemical structures of PIMs are shown in Figure 2A.

The above knowledge gaps exist because experimental techniques are limited in their ability to investigate PIM/LM/LAM and their associated biosynthesis and transport pathways. Traditional genetic techniques have uncovered many of the known enzymes in the PIM/LM/LAM pathway. However, mycobacteria are particularly challenging to investigate due to genetic redundancy among transport systems and the use of unique genes to synthesize and transport specialized substrates, limiting the ability to identify these genes from sequence similarity searches.¹⁰ In addition, traditional techniques do not provide a way to track non-genetically encoded glycans and lipids in cellular contexts. A powerful and complementary tool is metabolic labeling with unnatural carbohydrates bearing bioorthogonal handles that can be selectively modified via chemoselective “click” reactions, an approach pioneered by the Bertozzi group.¹¹ This approach has been applied to develop metabolic labeling tools for various bacterial cell envelope components, which has led to applications in whole-cell imaging, inhibitor development, cell envelope protein discovery, and the creation of novel diagnostic and therapeutic strategies.¹² Synthetic unnatural biosynthetic precursors have been developed to metabolically label an array of mycobacterial cell envelope components, including trehalose- and arabinogalactan-linked mycolates,^13–25 peptidoglycan,^26–28 and arabinogalactan.²⁹ No such probes have been developed for PIMs or the PIM-anchored derivatives LM and LAM in mycobacteria. The ability to metabolically incorporate inositol-based probes, such as azido inositol (InoAz) analogues, would yield a single substitution in the glycan core of PIMs (or additionally in terminal phospho-inositol capping residues of LAM in relevant organisms) and, in conjunction with bioorthogonal chemistry, would enable researchers to address many of the questions about PIM, LM, and LAM biosynthesis, transport, and dynamics raised above.

Herein, we report the development of InoAz analogues as metabolic labeling probes of PIM, LM, and LAM in the model organism Mycobacterium smegmatis (Msmeg). We found that 5-position-substituted inositol analogues, including 5-InoAz, are pathway-privileged for incorporation into early PIM species, LM, and LAM, thus providing a new tool to investigate these glycolipids in bacterial cells. Furthermore, in an effort to improve inositol analogue utilization by Msmeg cells, we leveraged 5-InoAz to generate a strain of Msmeg harboring mutations in an inositol utilization pathway, which led to the discovery of a previously unknown inositol importer and has implications for enhancing the incorporation of inositol analogues into PIM, LM, and LAM using genetic engineering. Thus, in addition to reporting a new tool for labeling PIM, LM, and LAM, this study highlights the utility of strain evolution approaches toward sugar auxotroph strains to identify spontaneous suppressor mutants that permit or enhance incorporation of azido sugars. Importantly, the ability to metabolically incorporate InoAz probes into PIM, LM, and LAM in live Msmeg and perform click chemistry-mediated detection of these species will open the door to addressing key knowledge gaps related to glycolipid biosynthesis and dynamics in mycobacteria.

Results and Discussion

Design and synthesis of InoAz analogues.

In considering the possible approaches to developing a metabolic labeling probe for PIM, LM, and LAM, we focused on inositol due to its presence in the structural core of all three glycolipids and the initiating role it plays in their biosynthesis (Figure 1). In Msmeg and M. tuberculosis, it has previously been shown that exogenously supplied inositol can support the growth of inositol auxotrophs lacking ino1, suggesting the presence of (a) pathway(s) in these organisms that can potentially be exploited for uptake and metabolic incorporation of InoAz analogues.³⁰ Because PIM, LM, and LAM are generally unmodified at the 3-, 4-, and 5-positions of the core inositol residue (with the exception of variable acylation at the 3-position), we designed 3-, 4-, and 5-InoAz as probe candidates (Figures 2A and 2B). We anticipated that subtle modification of these positions with small azido groups bearing native stereochemistry might be tolerated by endogenous PIM, LM, and LAM biosynthesis and transport machinery, while the azido group would enable subsequent modification of labeled glycans using azide-specific bioorthogonal chemistries. We envisioned that the InoAz probes would potentially be incorporated into the core of PIM, LM, and LAM, and/or the inositol phosphate cap appended to terminal arabinofuranosyl residues of LAM, a modification that occurs in Msmeg.¹ To access the target compounds, we used our recently reported Ferrier carbocyclization-mediated synthesis of enantiopure InoAz analogues,³¹ which was used here to produce 3- and 4-InoAz. Since this method failed to produce 5-InoAz due to an unusual azido group-ejecting elimination side reaction,³¹ we used Sureshan’s reported regioselective double S_N2 method to generate 5-InoAz.³²

Figure 2. InoAz incorporation into PIM/LM/LAM.

(A) Chemical structures of PIMs, LM, LAM. Box, structure of inositol phosphate cap modifying LAM terminal arabinofuranosyl residues in Msmeg. (B) Clickable azido inositol (InoAz) analogues and their use as metabolic labeling probes for PIM, LM, and LAM. (C) InoAz incorporation into PIM/LM/LAM and potential knock-out (KO) and overexpression (OE) genetic engineering strategies to favor probe incorporation are shown. *Indicates that this importer was discovered using 5-InoAz in the present work.

InoAz analogues do not support the growth of an inositol auxotroph.

As noted above, prior studies revealed the essentiality of inositol and some of its metabolic products, including PI and mono- and dimannosylated forms of PIMs,^33–35 in inositol auxotroph strains of mycobacteria disrupted in de novo inositol production, as well as the ability of exogenously supplied inositol to restore activity.^{36, 37} Therefore, we reasoned that an Msmeg inositol auxotroph (MsmegΔino1) would be a good model system to test the ability of our InoAz panel to be incorporated into Msmeg, and PIMs in particular, using a simple growth assay. Furthermore, we reasoned that the lack of competition from de novo-generated inositol in MsmegΔino1 would be advantageous for InoAz incorporation (Figure 2C). To assess growth rescue, MsmegΔino1 was grown to saturation with 5 mM inositol, washed to remove residual inositol with inositol-free medium, and then outgrown in inositol-free medium to deplete available inositol in the cytoplasm. Subsequently, the MsmegΔino1 cells were diluted and supplemented with InoAz analogues and growth was monitored. We observed growth of MsmegΔino1 in the presence of exogenously supplied inositol but not in the absence of inositol nor in the presence of 3-, 4-, or 5-InoAz (Figure S1). In addition, 3-, 4-, or 5-InoAz up to 10 mM concentration did not inhibit wild-type Msmeg growth (data not shown). These results suggested an inability of InoAz analogues to be substituted in place of inositol with high efficiency, due to either a lack of compound import or a downstream bottleneck in the PIM production pathway.

InoAz analogues differentially label an inositol auxotroph.

The inability of InoAz analogues to rescue the growth of the MsmegΔino1 strain suggests that the analogues are unable to substitute for native inositol in its metabolic pathways at a level sufficient to support growth. However, the InoAz analogues could be undergoing incorporation into glycolipids at lower levels not sufficient for growth rescue, and/or they may be processed early in the pathway by some enzymes before reaching a bottleneck in incorporation. To investigate these possibilities, we next evaluated the extent of InoAz incorporation into the cell envelope using fluorescent labeling and flow cytometry. Similar to growth assays, wild-type Msmeg or inositol auxotroph MsmegΔino1 cells were grown to saturation with 5 mM inositol, then washed into inositol-free medium and starved prior to treatment with InoAz analogues and subsequent labeling with AlexaFluor647 (AF647)-conjugated dibenzocyclooctyne (DBCO-AF647) via a strain-promoted azide−alkyne cycloaddition (SPAAC) reaction. Analysis of fixed cells by flow cytometry revealed that the three InoAz analogues differentially labeled the wild-type and inositol auxotroph strains. 3-InoAz did not measurably incorporate into either strain, whereas 4-InoAz equally labeled both wild-type Msmeg and MsmegΔino1 (Figure S2). In contrast, whereas 5-InoAz did not incorporate into wild-type Msmeg, it efficiently incorporated into the inositol auxotroph MsmegΔino1 when administered at 1 mM concentration (Figure 3A). Although the results from 4-InoAz warrant further investigation in the future, the ability of 5-InoAz to efficiently and selectively label MsmegΔino1, which is indicative of on-target incorporation via inositol metabolism, prompted us to focus on further defining its labeling characteristics in Msmeg.

Figure 3. 5-InoAz metabolically labels PIMs, LM, and LAM in an Msmeg inositol auxotroph.

(A) Msmeg wild-type or auxotroph strain MsmegΔino1 were treated 1 mM inositol (ctrl) or 5-InoAz, reacted with DBCO-AF647 via SPAAC, and analyzed by flow cytometry. MFI, Mean Fluorescence Intensity. Error bars represent the standard deviation of three replicate experiments. See Figure S2 in the Supporting Information for 3- and 4-InoAz labeling results. (B) MsmegΔino1,mshA::Tn5 was treated with 1 mM inositol or 5-InoAz in 7H9-ADC-tyloxapol medium at 37°C for 12 to 16 hours as described under Methods, then PI and PIMs were extracted and directly analyzed by LC-MS. Data shown are for Ac₁PIM₁ (Ac₁PIM₁ (C₅₁) exact mass = 1251.85; Ac₁PIM₁-Az (C₅₁) exact mass = 1276.86). See Figure S3 and Table S1 in the Supporting Information for complete LC-MS results. (C) MsmegΔino1,mshA::Tn5 was untreated (lane 1) or treated with 1 mM inositol (lane 2) or 5-InoAz (lane 3) as described in (B), then LM and LAM were extracted and analyzed by SDS-PAGE using either periodic acid–Schiff silver stain (PAS) (left) or reaction with DBCO-AF647 via SPAAC and fluorescence scanning (right).

5-InoAz incorporates into inositol-containing glycolipids of mycobacteria.

To directly examine 5-InoAz incorporation into PI and PIMs, 5-InoAz-treated Msmeg samples were subjected to glycolipid isolation and analysis by LC-MS. Note that these experiments were done in a strain we constructed to further favor InoAz incorporation, MsmegΔino1,mshA::Tn5, which is an inositol auxotroph and furthermore lacks the ability to channel exogenous inositol into the major intracellular reducing thiol in mycobacteria, mycothiol (Figure 2C). Using standard extraction procedures, total lipids (PIs and PIMs) were isolated from MsmegΔino1,mshA::Tn5 treated with either inositol as a control or 5-InoAz as described above for growth and flow cytometry assays. Extracted lipids were analyzed by LC-MS in negative ion mode for the presence of azide-substituted PI and PIMs. Promisingly, in lipid extracts of cells treated with 5-InoAz, we observed peaks corresponding to azide-substituted PI (PI-Az), PIM₁ (PIM₁-Az), and Ac₁PIM₁ (Ac₁PIM₁-Az), directly confirming incorporation of this probe into early PIM species (Figures 3B and S3, Supporting Information). However, 5-InoAz labeling was not detected in PIM₂, Ac₁PIM₂, or higher PIMs with additional mannosylation, suggesting either inefficient or no labeling of these species (Figure S3, Supporting Information), which is consistent with the inability of 5-InoAz to rescue growth of MsmegΔino1 (Figure S1). LC-MS analysis also revealed altered distribution of PIMs in 5-InoAz-treated versus inositol-treated cells, including accumulation of PIM₁ forms and absence of PIM₂ forms (Table S1), suggesting a possible bottleneck in the generation of PIM₂ species by the PimB’ mannosyltransferase (MSMEG_4253) (see Figure 1). In PIM species, the inositol residue is modified at the 1- (phosphoester), 2- (mannose), and 6- (mannose) positions. It is possible that azido group substitution in place of the hydroxyl group at the inositol 5-position disrupts binding to or processing by the PimB’ mannosyltransferase at the C6 hydroxyl, but not by upstream PimA at the C2 position. This finding provides insights for future engineering strategies to promote incorporation of 5-InoAz into higher PIM species, such as overexpression or active site engineering of PimB’.

We next used SDS-PAGE to investigate whether 5-InoAz was incorporated into the highly glycosylated PIM derivatives LM and LAM. By SDS-PAGE, LM and LAM separate into two distinct molecular weight fractions that can be detected by periodic acid–Schiff silver stain (PAS), whereas azide labeling can be detected by click chemistry. MsmegΔino1,mshA::Tn5 was treated with 1 mM inositol or 5-InoAz, or left untreated, then LM and LAM were extracted using a standard procedure and subjected to SPAAC reaction with DBCO-AF647 before separation by SDS-PAGE and visualization by in-gel fluorescence scanning. Fluorescence images were then compared to the same gel after PAS staining. LM and LAM extracted from 5-InoAz-treated cells displayed robust fluorescence corresponding to expected LM and LAM bands, whereas unlabeled controls had no detectable signal (Figure 3C). The PAS-stained gel showed that each sample had equivalent amounts of LM and LAM material, confirming that the fluorescence signal was azide-specific. In addition, we confirmed that the signal did not originate from unincorporated InoAz that retained in the gel, as the cycloaddition product of DBCO-AF647 incubated with InoAz monosaccharide did not migrate into the LM or LAM regions and instead migrated at the dye front (data not shown).

Together, our LC-MS and SDS-PAGE data demonstrate that, in addition to labeling PIMs up to Ac₁PIM₁, 5-InoAz incorporates into the PIM-anchored glycolipids LM and LAM. Given the apparent bottleneck of 5-InoAz labeling at Ac₁PIM₁ described above, the observed azide-labeling of LM and LAM could be explained by a low level of core inositol labeling, inositol phosphate cap labeling, or both. To provide insight into this issue, we subjected 5-InoAz-labeled LM and LAM extracts to DBCO-AF647 reaction followed by digestion with Cellulomonas gelida endoarabinanase, which cleaves internal glycosidic bonds in the LAM arabinan domain, thereby leading to the loss of inositol phosphate-capped arabinan termini.³⁸ SDS-PAGE analysis of these samples showed that endoarabinanase digestion led to loss of intact LAM and diminished total in-gel fluorescence intensity by ~60%, which is consistent with appreciable incorporation of 5-InoAz into the inositol phosphate LAM capping groups (Figure S4). However, following endoarabinanase treatment, ~50% of the fluorescence was still retained in the LM-like band (i.e., endoarabinanase-digested LAM), which, along with the signal in LM from undigested extracts, is indicative of some core labeling (Figure S4). Overall, our data are consistent with a scenario in which 5-InoAz labels PIMs, LM, and LAM in MsmegΔino1,mshA::Tn5, with lower PIMs (up to Ac₁PIM₁) being efficiently labeled in the structural core, higher PIM/LM/LAM species having diminished core labeling due to a bottleneck at PimB’ required to make PIM₂/Ac₁PIM₂, and LAM being additionally labeled in its inositol phosphate cap. The absence of detectable 5-InoAz-labeled or -unlabeled PIM₂, Ac₁PIM₂, or higher forms of PIMs in 5-InoAz-treated cells while LM and LAM are produced suggests that when inositol incorporation is limited, Msmeg favors the synthesis of mature LM and LAM (elongation products of PIM₂/Ac₁PIM₂) over that of intermediate PIM species.

Isolation and sequencing of spontaneous suppressor mutants that allow growth on 5-InoAz.

As a potential strategy to overcome bottlenecks associated with 5-InoAz incorporation, we reasoned that extended growth in the presence of 5-InoAz as a singular source of inositol could generate an evolved strain of Msmeg able to utilize 5-InoAz more efficiently. As we and others have observed, the metabolic products of native inositol, including PIMs, are required for mycobacterial growth. Therefore, MsmegΔino1,mshA::Tn5 is under strong selective pressure to incorporate 5-InoAz into PIMs. The inability of 5-InoAz-treated cells to replicate (Figure S1) provides a simple assay to select strains that have acquired advantageous mutations for growth. MsmegΔino1,mshA::Tn5 was grown to saturation in inositol-containing medium, and then washed and outgrown in inositol-free medium. Cells were then diluted and subjected to extended growth in 7H9 medium that was either supplemented with 1 mM 5-InoAz as the sole source of inositol, or left untreated. Whereas cells that received no additive continued to decline, the culture supplemented with 5-InoAz increased exponentially after a 2-day outgrowth period, suggesting an acquired ability to utilize 5-InoAz (Figure 4A). After 8 days, the 5-InoAz-treated culture was dilution-plated on solid medium containing 1 mM 5-InoAz. Single colonies of ten selected clones were then directly grown in 7H9 liquid medium containing 1 mM 5-InoAz without the need for native inositol, showing that these strains had evolved (a) mechanism(s) that permitted efficient utilization of 5-InoAz as a surrogate for inositol.

Figure 4. 5-InoAz-derived spontaneous suppressor mutant reveals an inositol utilization pathway in Msmeg.

(A) MsmegΔino1,mshA::Tn5 was grown with no additive, 5 mM inositol, or 5 mM 5-InoAz and monitored by optical density at 600 nm over 192 hours. (B) Genetic organization of a potential inositol utilization gene cluster in Msmeg. (*) Denotes gene containing mutations observed by whole-genome sequencing. (C) CRISPRi knockdown of Msmeg genes of interest, MSMEG_4657 and MSMEG_4659 in wild-type Msmeg and MsmegΔino1. Strains were transformed with either empty plasmid or plasmid containing a gene-specific guide RNA sequence. Strains were serially diluted on plates without or with ATc (0 or 100 ng/mL) and varied amounts of supplemental inositol (0, 0.1, or 1 mM). Conditional knock-down (cKD) of MSMEG_4659 supports growth on ten-fold less inositol than parent strain alone and cKD of MSMEG_4657 abrogates growth of auxotroph. (D) MsmegΔino1,mshA::Tn5 without and with expression of the ABC transporter MSMEG_4656–4658 was monitored for growth on various amounts of inositol (0, 0.5, and 5 mM) by optical density at 600 nm.

To identify the mutations responsible for permitting growth on 5-InoAz, we performed whole genome sequencing (WGS) of 4 selected suppressor mutant clones from the 5-InoAz-treated culture. WGS revealed independent non-synonymous mutations in MSMEG_4659, which were confirmed by Sanger sequencing in 6 other strains (Table S2). All strains harbored frameshift mutations (nucleotide C724insC, C480del, C252insC, C348insA, A270insA) or introduced STOP codons into the MSMEG_4659 gene sequence (nucleotide C67T, C235T), with the exception of one isolated clone that harbored a non-synonymous mutation (nucleotide C233T, T>I) in the same gene. All sequenced clones possessed additional mutations in different genes compared to parental control strains, as indicated in Figure S5.

Further characterization efforts were focused on the common mutated gene, MSMEG_4659. MSMEG_4659 is predicted to encode a GntR-like bacterial transcription factor (PFAM: PF00392). These transcription factors generally possess an N-terminal DNA-binding domain and a C-terminal oligomerization or effector domain.³⁹ Furthermore, transcription factors of the GntR family are known to function on carbohydrate metabolism genes.⁴⁰ A sequence similarity search to identify the nearest potential homolog of MSMEG_4659 revealed an intriguing link to a transcription factor IolR, which is involved in the control of inositol catabolism within soil bacteria and Corynebacterium glutamicum, a close relative of Mycobacterium that possesses PIM, LM, and LAM⁴¹ (IolR, Cgl1057 (cg0196), 55% shared nucleotide identity to MSMEG_4659).⁴² Our examination of genes immediately downstream of MSMEG_4659 (Figure 4B) revealed a cluster of 6 genes (MSMEG_4661–MSMEG_4666) with sequence similarity to genes reported to catabolize inositol to acetyl-CoA and dihydroxyacetone in various soil bacteria.⁴³ Reports investigating genes under control of IolR in Corynebacterium glutamicum⁴² also revealed genes involved in inositol importation. Notable in Msmeg is the presence of an ABC transporter gene cluster immediately upstream of MSMEG_4659, MSMEG_4656–4658 (Figure 4B). The ability of mycobacteria to import inositol has previously been demonstrated from experiments involving inositol auxotrophs^{30, 36} and from studies with radiolabeled inositol.⁴⁴ Here, we found that wild-type Msmeg can grow on inositol (but not 5-InoAz) as a sole carbon source, confirming its ability to import native inositol (Figure S6). However, putative transporter genes have only been suggested from bioinformatic analyses for mycobacterial species, and MSMEG_4656–4658 was not identified as a candidate importer in these studies.^{45, 46}

Silencing MSMEG_4659 and MSMEG_4656–4658 gene expression reveals roles in inositol utilization.

To investigate whether the GntR transcriptional repressor MSMEG_4659 and the putative ABC-transporter encoded by MSMEG_4656–4658 played a role in inositol importation or utilization, we used our inositol auxotroph, MsmegΔino1, to study the effect of silencing these genes on Msmeg growth. Since MsmegΔino1 is reliant on the import and utilization of exogenous inositol for growth, silencing genes critical to these functions was expected to lead to the growth arrest of this auxotroph.

We used CRISPR interference (CRISPRi) methodology⁴⁷ to knock-down gene expression of MSMEG_4659 and MSMEG_4657, the gene encoding the transmembrane subunit of the adjacent ABC transporter. To silence MSMEG_4659 and MSMEG_4657, plasmids encoding guide RNAs to target these genes were designed according to reported PAM strength guidelines⁴⁷ (Table S3) and transformed into MsmegΔino1 and wild-type Msmeg as a control. MsmegΔino1 strains containing CRISPRi plasmids were maintained in the presence of inositol and no inducer, then washed in inositol-free medium and plated on solid medium with or without inositol (0, 0.1, and 1 mM) and with or without inducer (anhydrotetracycline (ATc); 0 or 100 ng/mL to allow for gene expression or repression, respectively) then incubated alongside wild-type control strains containing CRISPRi plasmids and empty plasmids. As expected, wild-type Msmeg grew under all conditions, whereas MsmegΔino1 only grew when supplemented with 1 mM inositol. When MSMEG_4659 was knocked down in MsmegΔino1 (MsmegΔino1(MSMEG_4659(−))), the resulting strain was able to grow on ten-fold less inositol (0.1 mM) (Figure 4C), most likely the result of MSMEG_4659 silencing leading to loss of repression and subsequent enhanced expression of genes involved in inositol importation and/or utilization. In contrast, when MSMEG_4657, the gene encoding the transmembrane protein of the MSMEG_4656–4658 ABC transporter, was silenced in MsmegΔino1 (MsmegΔino1(MSMEG_4657(−))), there was a lack of growth on 1 mM inositol (Figure 4C), likely due to an inability of the strain to import exogenous inositol. We also designed guide RNAs to target transporters suggested by the literature based on their similarity to known inositol importer genes, including MSMEG_5166, MSMEG_0190, MSMEG_5161, MSMEG_5559, and MSMEG_4462. We found that when silenced, these genes did not prevent growth on 1 mM inositol in the presence of ATc, indicating that these genes are unlikely to be directly responsible for inositol importation (Figure S7).

Finally, to gain better insight into the function of the ABC transporter MSMEG_4656–4658, we overexpressed the three encoding genes from a replicative plasmid under control of the hsp60 promoter and monitored growth at 600 nm as a function of exogenously supplied inositol concentrations. We found that the inositol auxotroph strain MsmegΔino1,mshA::Tn5 overexpressing the ABC transporter MSMEG_4656–4658 grew more robustly on inositol and exhibited growth at lower concentrations of inositol than the auxotroph alone (Figures 4C and 4D). Combined with the CRISPRi silencing experiments, these results strongly support that MSMEG_4656–4658 is the sole transporter responsible for inositol importation in Msmeg.

Conclusion

Inositol-containing glycans in mycobacteria, including PIM, LM, and LAM, have drawn significant attention due to their important roles in the physiology and pathogenesis of multiple human pathogens.⁴⁵ Yet, many questions remain regarding the PIM/LM/LAM pathway, including the identity of many proteins involved in their biosynthesis and transport, as well as their dynamics in live cells and during infection. Here, we reported the development and characterization of synthetic InoAz analogues, in particular 5-InoAz, as the first tools for metabolic labeling of these glycolipids. Given the availability of various azide-specific bioorthogonal chemistries and associated strategies for the visualization and enrichment of azide-labeled biomolecules and their interactors,^{11, 12} we anticipate that InoAz analogues will open myriad new avenues to investigate the as-yet incompletely characterized PIM/LM/LAM pathway.

In the present work, our screen of three regioisomeric InoAz analogues revealed that the azido group position had a critical impact on metabolic incorporation, which is consistent with previous findings on azide-modified substrates for labeling of trehalose mycolates¹⁴ and arabinogalactan.²⁹ Our current study mainly focused on the 5-InoAz isomer, which incorporated into early PIMs, LM, and LAM in Msmeg strains that were rationally engineered to favor probe incorporation. Furthermore, we took an unbiased genetic approach to improving inositol analogue uptake by growing an inositol auxotroph on 5-InoAz over an extended period, which selected for strains with mutations that enhanced utilization of 5-InoAz. Characterization of these “evolution” strains led to the discovery of a previously unknown ABC-transporter for inositol import in Msmeg, MSMEG_4656–4658 which we here propose to rename InoABC, which represents the first inositol importer identified in mycobacteria. This finding has several implications. First, it highlights the ability of azido sugar probes to elucidate biology through spontaneous mutation approaches, as also recently demonstrated by the use of 6-azido trehalose inhibitory activity to identify PPE51 as the protein required for trehalose uptake across the mycomembrane in M. tuberculosis.⁴⁸ Second, it suggests next steps for further improving InoAz labeling of PIM, LM, and LAM, including generation of “evolution” strains for other InoAz isomers, as well as overexpressing the inositol importer to favor InoAz uptake. Third, it opens up new avenues for research on mycobacterial inositol metabolism, in particular determining the role of inositol import in mycobacterial physiology and, potentially, virulence.

There are a number of other future directions for the development, characterization, and application of metabolic labeling probes for PIM, LM, and LAM. Herein, we reported both rational and unbiased genetic approaches to improving InoAz incorporation into glycolipids. Additional approaches, such as overexpression or active site engineering of bottleneck enzymes (i.e., PimB’), could also be beneficial. On the other hand, the requirement to genetically alter mycobacteria to enable probe usage can be a limitation, especially with respect to extending the tools to other mycobacterial species. In this regard, it was promising to find that 4-InoAz labeled wild-type Msmeg, which we are presently investigating in more detail. In addition, other metabolic precursors to PIM/LM/LAM, such as mannose derivatives or, in the case of LAM, arabinose derivatives, could be pursued as complementary probes for these glycolipids. Additional inositol probe improvement may be achieved through the use of alternative labeling handles (e.g., alkynes), inclusion of extended linkers, and/or a probe design that maintains potential H-bonding interactions, as previously used in the labeling of eukaryotic glycosylphosphatidylinositol (GPI) anchors.⁴⁹ In this study, we also discussed the potential of InoAz probes to label the inositol core of PIM/LM/LAM, the inositol phosphate cap of LAM, or both. All of these outcomes are useful, as they enable investigation of complementary aspects of PIM/LM/LAM biosynthesis and export. Overall, novel methods to specifically label PIMs and PIM-derived LM and LAM are expected to be a valuable new addition to the expanding toolbox of mycobacterial cell envelope probes.

Experimental Procedures

Strains, culture media, and reagents.

M. smegmatis mc²155 (Msmeg) was used as the parental strain, and the mutants derived from it were routinely grown in 7H9 broth supplemented with 0.5% (w/v) glycerol and 10% (v/v) albumin–dextrose–catalase (ADC) and 0.05% tyloxapol at 37°C. The same medium with the addition of 1.5% (w/v) agar was used as solid medium. Escherichia coli strain DH5α was maintained in LB medium, propagated at 37°C, and used for routine cloning and transformation experiments. For wash steps, 1X phosphate-buffered saline (PBS, Fisher) was used. Antibiotics, when required, were added at the following concentrations: for E. coli, streptomycin 50 μg ml⁻¹, kanamycin 50 μg ml⁻¹; for Msmeg, streptomycin 20 μg ml⁻¹, kanamycin 20 μg ml⁻¹, hygromycin 50 μg ml⁻¹, anhydro-tetracycline (ATc) 100 ng ml⁻¹. For supplementation with inositol, a high millimolar stock (200–250 mM) of inositol (Sigma) or InoAz was prepared in water. DBCO-AF647 (click chemistry tools) was dissolved in DMSO as a 1 mM stock.

Chemical Synthesis.

3- and 4-InoAz were synthesized as reported by Ausmus et al.³¹ 5-InoAz was synthesized as reported by Ravi et al.^{32 1}H and ¹³C NMR spectral data for intermediates and products were acquired on either Varian Mercury 300, Varian Inova 500, or Bruker Avance Neo 500 systems, and matched the literature.

Genetic Manipulations.

The construction of an Msmeg mc²155 strain that is an auxotroph for inositol involved replacement of the corresponding open reading frame (Ino1: MSMEG_6904) with a streptomycin resistance cassette from pHP45Ω following standard allelic replacement strategies with pPR27-xylE, a temperature-sensitive replicative plasmid containing the counter selectable marker sacB, and colored marker xylE.^{50, 51} Briefly, ~500 bp upstream and downstream fragments from MSMEG_6904 were amplified from primers (Table S3) and cloned into shuttle vector pMV261 via isothermal annealing cloning, then the entire cloned segment was digested and ligated into pPR27-xylE generating the plasmid pPR27-xylE-Ino1. This construct was then introduced via electroporation into Msmeg mc²155, or an mshA-interrupted Msmeg mc²155 strain containing a Tn5 insertion at MSMEG_0933 reported previously.⁵² Transformants were selected on 7H9 ADC agar plates supplemented with streptomycin (50 μg ml−1) at 30°C. Colonies transformed with pPR27-xylE-ino1 were detected by index plating and development with 1% catechol. pPR27-xylE-ino1 colonies were patched onto 7H9 ADC agar plates containing sucrose (10% w/v), streptomycin (50 μg ml⁻¹), and inositol (5 mM), followed by incubation at 42°C. Insertional mutant colonies were screened for loss of pPR27-xylE-ino1 delivery plasmid by selecting colonies that remained white when sprayed with catechol. These candidate deletion colonies were further validated by replica plating on 7H9 ADC agar plates supplemented with appropriate antibiotics and with and without inositol. Both single (ino1: ΔMSMEG_6904) and double (ino1 and mshA: ΔMSMEG_0933-ΔMSMEG_6904) mutants were confirmed by PCR using specific primer combinations for MSMEG_6904 (Table S3) and by whole-genome sequencing.

For CRISPRi plasmids preparation, briefly, PAM sequences were chosen according to predicted strength and sequentially searched in the coding sequence of the gene of interest. Subsequently, sgRNA targeting sequences were designed with two unique BsmBI restriction sites and ligated in CRISPRi plasmid pLJR962.⁴⁷

Metabolic Labeling.

For metabolic labeling, cells were washed extensively to remove all residual inositol as indicated in the results section and subsequently cultured in 7H9-ADC-tyloxapol (0.05%) medium supplemented with 1 mM of the indicated InoAz analogue and incubated at 37°C for 12–16 h.

Fluorescent labeling.

After metabolic labeling, cells were washed at least 3 times in PBS and resuspended at an optical density of 0.2–0.5 in PBS with 20 μM DBCO-AF647 (1:50, from a 1 mM stock in DMSO) for 30 min at room temperature on a tube rotator in the dark. Fluorescently labeled cells were then washed 3 times in PBS to remove residual dye.

Flow cytometry.

Following metabolic and fluorescent labeling, cells were harvested and washed with 1x PBS and transferred to 5 mL polystyrene tubes. Flow cytometry was performed on a three-laser Cytek Aurora flow cytometer. Fluorescence data was collected for 50,000 events, and Flow Cytometry Standard (FCS) file data were analyzed using FlowJo software (BD Biosciences). Gates were set first to reduce scatter and then to select singlets. Histogram data were transformed to calculate median fluorescence intensity.

PIM, LM and LAM extraction and analysis.

Extraction of PIMs, LM and LAM, digestion of LAM with Cellulomonas gelida endoarabinanase and analysis on intact and digested glycolipid and lipoglycan products by TLC, SDS-PAGE and LC/MS followed earlier procedures.^53–57 For samples digested by endoarabinanase,^{58, 59} dried samples were resuspended in purified endoarabinanase and incubated overnight in a water bath at 37°C. Samples were then dried in a speed-vac and fluorescently labeled and separated on SDS-PAGE as described above.

Whole genome sequencing.

Genomic DNA was extracted from mid-logarithmic phase bacteria growing in 7H9 media using Qiagen UCP Pathogen kit (Qiagen). Illumina libraries were prepared from 830–931 ng of mechanically fragmented DNA (300 bp, Covaris M instrument per manufacturer protocol) using the Kapa Hyper prep kit and quantified using the Qubit double-stranded DNA (dsDNA) BR assay kit (Thermo Fisher Scientific). Fragment size was assessed on a fragment analyzer (Advanced Analytical Technologies). Libraries were multiplexed and sequenced as 75-base-long single-end reads on an Illumina NextSeq 500 instrument. Reads were adapted and quality trimmed with Trimmomatic v0.33 with the following threshold: quality threshold of 15 and minimum length before dropping reads of 40. Trimmed reads were then mapped onto the Msmeg mc²155 reference genome (RefSeq NC_008596.1) using Bowtie2 v2.2.5. Variant calling was performed using Varscan (min coverage: 5; min reads: 5; average quality: 15; min variant allele frequency: 0.01; min for homoplasy: 0.9). Vcf were merged with bcftools option –merge.

Data availability.

The sequencing data described in this publication have been submitted to the NCBI Gene Expression Omnibus (GEO) under BioProject #PRJNA913569.